Imperfect diagnostic automation: How adjusting bias and saliency affects operator trust.

摘要

We report three experiments that examine the effects of imperfect automation on operator trust and dependence. By manipulating the automation bias to either present false alarms or misses, the data reveal patterns of operator behavior consistent with a multiple-process cognitive model that allows non-selective effects of automation errors on operator behavior. These non-selective effects were eliminated by reducing the salience of the automation errors.